CN107358240A - A kind of object detecting method and system by feature templates construction feature pond - Google Patents
A kind of object detecting method and system by feature templates construction feature pond Download PDFInfo
- Publication number
- CN107358240A CN107358240A CN201710426358.4A CN201710426358A CN107358240A CN 107358240 A CN107358240 A CN 107358240A CN 201710426358 A CN201710426358 A CN 201710426358A CN 107358240 A CN107358240 A CN 107358240A
- Authority
- CN
- China
- Prior art keywords
- feature
- image
- object detection
- templates
- histogram
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/469—Contour-based spatial representations, e.g. vector-coding
- G06V10/473—Contour-based spatial representations, e.g. vector-coding using gradient analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of object detecting method and system by feature templates construction feature pond, it is by handling original image, obtain gray-scale map, gradient magnitude figure, histogram of gradients, and the feature templates of more than one different grids are set, and by the feature templates respectively in the gray-scale map, gradient magnitude figure, moved in histogram of gradients, obtain the feature pool of characteristics of image, the characteristics of image in the feature pool is trained using adaboost algorithms to obtain optimal object detection feature, object detection is finally carried out according to the optimal object detection feature;The present invention carries out extracting characteristics of image and construction feature pond by using different characteristic template to heterogeneity image, the more features of image can be obtained, fully to excavate the useful information of image and different objects clarification of objective can be distinguished, to select to obtain optimal object detection feature from the feature pool, detection efficiency and accuracy rate are improved.
Description
Technical field
The present invention relates to image analysis technology field, particularly a kind of object detection by feature templates construction feature pond
The system of method and its application this method.
Background technology
Target detection technique is application critically important in image domains, in intelligent monitoring, traffic block port, recognition of face, figure
It is very wide as being applied in retrieval.In addition to traditional image processing method detection specific objective, trained by the method for machine learning
The model of specific objective, in the picture gliding model window are come to detect target be very effective method up to now, especially
For rigid objects.
In using method of the machine learning method to detect specific objective, SVM (Support Vector Machine)
It is method the most frequently used in image object detection with Adaboost algorithm.Wherein, the Haar features that Viola and Jones is proposed+
Adaboost methods have preferable Detection results, especially in terms of Face datection.However, haar characteristic recognition methods are main
Suitable for having the object of obvious, rock-steady structure haar features as face is this, and detected for pedestrian detection, car tracing
Deng ineffective on moving object segmentation.
The content of the invention
The present invention to solve the above problems, provide a kind of object detecting method by feature templates construction feature pond and
System, it carries out extracting characteristics of image and construction feature pond to heterogeneity image by using different characteristic template, can obtained
To the more features of image, fully to excavate the useful information of image and different objects clarification of objective can be distinguished, so as to from institute
State and select to obtain optimal object detection feature in feature pool, improve detection efficiency and accuracy rate.
To achieve the above object, the technical solution adopted by the present invention is:
A kind of object detecting method by feature templates construction feature pond, it comprises the following steps:
(10) original image is handled, obtains gray-scale map, gradient magnitude figure, histogram of gradients;
(20) feature templates of more than one different grids are set, and by the feature templates respectively the gray-scale map,
Moved on gradient magnitude figure, histogram of gradients, obtain the feature pool of characteristics of image;
(30) characteristics of image in the feature pool is trained using adaboost algorithms to obtain optimal object inspection
Survey feature;
(40) object detection is carried out according to the optimal object detection feature.
Preferably, in described step (10), the histogram of gradients is further averagely divided into 9 from 0 to 360 degree scope
Individual section, each section are 40 degree, obtain 9 histogram of gradients.
Preferably, in described step (20), the feature templates include 12 kinds, single according to the white of every kind of feature templates
The layout of first lattice and black unit lattice obtains the algorithm of 12 kinds of image feature values, by by 12 kinds of feature templates respectively in the ash
Moved in degree figure, gradient magnitude figure, histogram of gradients by default step-length, extract substantial amounts of characteristics of image, and calculated each
Characteristic value corresponding to characteristics of image.
Preferably, described image characteristic value is that the pixel value sum in the white cells lattice subtracts the black unit lattice
Interior pixel value sum.
Preferably, in described step (30), refer to the characteristic value and thing of each characteristics of image in the feature pool
The characteristic value of background is compared where body, and distinguishes maximum characteristics of image as most using with the characteristic value of background where object
Good object detection feature.
Preferably, in described step (40), refer to carry out construction physical examination using the optimal object detection feature
Grader is surveyed, object detection is carried out by the object detection classifier.
Accordingly, the present invention also provides a kind of object detecting system by feature templates construction feature pond, and it includes:
Image pre-processing module, for handling original image, obtain gray-scale map, gradient magnitude figure, gradient Nogata
Figure;
Feature pool builds module, its feature templates by setting more than one different grids, and by the feature templates
Moved respectively in the gray-scale map, gradient magnitude figure, histogram of gradients, obtain the feature pool of characteristics of image;
Features training module, it uses adaboost algorithms to be trained to obtain most to the characteristics of image in the feature pool
Good object detection feature;
Object detection module, it carries out object detection according to the optimal object detection feature.
Preferably, in described feature pool structure module, the feature templates include 12 kinds, according to every kind of feature templates
The layout of white cells lattice and black unit lattice obtains the algorithm of 12 kinds of image feature values, by the way that 12 kinds of feature templates are existed respectively
Moved in the gray-scale map, gradient magnitude figure, histogram of gradients by default step-length, extract substantial amounts of characteristics of image, and count
Calculate characteristic value corresponding to each characteristics of image.
Preferably, in described features training module, refer to the characteristic value of each characteristics of image in the feature pool
It is compared with the characteristic value of background where object, and makees maximum characteristics of image is distinguished with the characteristic value of background where object
For optimal object detection feature.
Preferably, in described object detection module, refer to carry out construction using the optimal object detection feature
Grader is surveyed in physical examination, and object detection is carried out by the object detection classifier.
The beneficial effects of the invention are as follows:
The present invention carries out extracting characteristics of image and construction feature pond by using different characteristic template to heterogeneity image,
The more features of image can be obtained, fully to excavate the useful information of image and different objects clarification of objective can be distinguished, with
Just select to obtain optimal object detection feature from the feature pool, improve detection efficiency and accuracy rate.
Brief description of the drawings
Accompanying drawing described herein is used for providing a further understanding of the present invention, forms the part of the present invention, this hair
Bright schematic description and description is used to explain the present invention, does not form inappropriate limitation of the present invention.In the accompanying drawings:
Fig. 1 is a kind of general flow chart of object detecting method by feature templates construction feature pond of the invention;
Fig. 2 is a kind of structural representation of object detecting system by feature templates construction feature pond of the invention;Fig. 3 is
The design diagram of 12 kinds of feature templates of one specific embodiment.
Embodiment
In order that technical problems, technical solutions and advantages to be solved are clearer, clear, tie below
Closing drawings and Examples, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only used
To explain the present invention, it is not intended to limit the present invention.
As shown in figure 1, a kind of object detecting method by feature templates construction feature pond of the present invention, it includes following
Step:
(10) original image is handled, obtains gray-scale map, gradient magnitude figure, histogram of gradients;
(20) feature templates of more than one different grids are set, and by the feature templates respectively the gray-scale map,
Moved on gradient magnitude figure, histogram of gradients, obtain the feature pool of characteristics of image;
(30) characteristics of image in the feature pool is trained using adaboost algorithms to obtain optimal object inspection
Survey feature;
(40) object detection is carried out according to the optimal object detection feature.
In described step (10), the histogram of gradients is further averagely divided into 9 sections from 0 to 360 degree scope,
Each section is 40 degree, obtains 9 histogram of gradients.That is, there are 11 heterogeneity images in the present embodiment, including:1 ash
Figure, 1 gradient magnitude figure, 9 histogram of gradients are spent, optimal effectiveness can be obtained.Can also be as needed to institute in practical application
State histogram of gradients and demarcation interval is carried out in the range of 0 to 180 degree, section quantity can also carry out increase and decrease setting as needed, no
As limit.
In described step (20), the feature templates include 12 kinds, 12 kinds of feature templates spy different equivalent to 12 kinds
Calculation is levied, 12 kinds of image feature values are obtained according to the layout of the white cells lattice of every kind of feature templates and black unit lattice
Algorithm, by the way that 12 kinds of feature templates are carried out in the gray-scale map, gradient magnitude figure, histogram of gradients by default step-length respectively
It is mobile, substantial amounts of characteristics of image is extracted, and calculate characteristic value corresponding to each characteristics of image.Described image characteristic value is described white
Pixel value sum in color element lattice subtracts the pixel value sum in the black unit lattice, the pixel in the black unit lattice
Value sum can be zero.;Wherein, the big I of the white cells lattice or black unit lattice is adjusted according to training pattern size
It is whole, such as may be configured as 3*3 or 6*6.
As shown in figure 3, in the present embodiment, 12 kinds of feature templates include:
1. only include a white cells lattice;
2. include a white cells lattice and a black unit lattice and the two lateral arrangement;
3. include a white cells lattice and a black unit lattice and the two is longitudinally arranged;
4. including four white cells lattice and two horizontal strokes, two vertical arrangement;
5. include laterally continuous two white cells lattice and laterally continuous two black unit lattice and the two longitudinal cloth
Put;
6. including being longitudinally continuous two white cells lattice and being longitudinally continuous two black unit lattice and the two horizontal cloth
Put;
7. include two white cells lattice and two black unit lattice and the two two horizontal stroke two is vertical diagonally arranged;
8. include laterally continuous two black unit lattice and a white cells lattice and the two lateral arrangement;
9. include a white cells lattice and laterally continuous two black unit lattice and the two lateral arrangement;
10. include laterally continuous three black unit lattice and laterally continuous three white cells lattice and the two longitudinal cloth
Put;
11. include two groups of the 8th kind of feature templates and the two is longitudinally arranged;
12. include two groups of the 9th kind of feature templates and the two is longitudinally arranged.
In described step (30), refer to carry on the back where the characteristic value of each characteristics of image in the feature pool and object
The characteristic value of scape is compared, and distinguishes maximum characteristics of image as optimal object using with the characteristic value of background where object
Detect feature.
In described step (40), refer to carry out building object detection classification using the optimal object detection feature
Device, object detection is carried out by the object detection classifier.
As shown in Fig. 2 corresponding with the object detecting method, the present invention also provides one kind and built by feature templates
The object detecting system of feature pool, it includes:
Image pre-processing module, for handling original image, obtain gray-scale map, gradient magnitude figure, gradient Nogata
Figure;
Feature pool builds module, its feature templates by setting more than one different grids, and by the feature templates
Moved respectively in the gray-scale map, gradient magnitude figure, histogram of gradients, obtain the feature pool of characteristics of image;
Features training module, it uses adaboost algorithms to be trained to obtain most to the characteristics of image in the feature pool
Good object detection feature;
Object detection module, it carries out object detection according to the optimal object detection feature.
In described feature pool structure module, the feature templates include 12 kinds, single according to the white of every kind of feature templates
The layout of first lattice and black unit lattice obtains the algorithm of 12 kinds of image feature values, by by 12 kinds of feature templates respectively in the ash
Moved in degree figure, gradient magnitude figure, histogram of gradients by default step-length, extract substantial amounts of characteristics of image, and calculated each
Characteristic value corresponding to characteristics of image.
In described features training module, refer to the characteristic value of each characteristics of image in the feature pool and object institute
It is compared in the characteristic value of background, and maximum characteristics of image is distinguished as most preferably using with the characteristic value of background where object
Object detection feature.
In described object detection module, refer to carry out building object detection point using the optimal object detection feature
Class device, object detection is carried out by the object detection classifier.
It should be noted that each embodiment in this specification is described by the way of progressive, each embodiment weight
Point explanation is all difference with other embodiment, between each embodiment identical similar part mutually referring to.
For system class embodiment, because it is substantially similar to embodiment of the method, so description is fairly simple, related part is joined
See the part explanation of embodiment of the method.
Also, herein, term " comprising ", "comprising" or its any other variant are intended to the bag of nonexcludability
Contain, so that process, method, article or equipment including a series of elements not only include those key elements, but also including
The other element being not expressly set out, or also include for this process, method, article or the intrinsic key element of equipment.
In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that including the key element
Process, method, other identical element also be present in article or equipment.In addition, those of ordinary skill in the art can manage
Solution realizes that all or part of step of above-described embodiment can be completed by hardware, can also instruct correlation by program
Hardware is completed, and described program can be stored in a kind of computer-readable recording medium, and storage medium mentioned above can be with
It is read-only storage, disk or CD etc..
The preferred embodiments of the present invention have shown and described in described above, it should be understood that the present invention is not limited to this paper institutes
The form of disclosure, the exclusion to other embodiment is not to be taken as, and can be used for various other combinations, modification and environment, and energy
Enough in this paper invented the scope of the idea, it is modified by the technology or knowledge of above-mentioned teaching or association area.And people from this area
The change and change that member is carried out do not depart from the spirit and scope of the present invention, then all should be in the protection of appended claims of the present invention
In the range of.
Claims (10)
1. a kind of object detecting method by feature templates construction feature pond, it is characterised in that comprise the following steps:
(10) original image is handled, obtains gray-scale map, gradient magnitude figure, histogram of gradients;
(20) feature templates of more than one different grids are set, and by the feature templates respectively in the gray-scale map, gradient
Moved on amplitude figure, histogram of gradients, obtain the feature pool of characteristics of image;
(30) characteristics of image in the feature pool is trained using adaboost algorithms to obtain optimal object detection spy
Sign;
(40) object detection is carried out according to the optimal object detection feature.
A kind of 2. object detecting method by feature templates construction feature pond according to claim 1, it is characterised in that:
In described step (10), the histogram of gradients is further averagely divided into 9 sections, each section from 0 to 360 degree scope
For 40 degree, 9 histogram of gradients are obtained.
3. a kind of object detecting method by feature templates construction feature pond according to claim 1 or 2, its feature exists
In:In described step (20), the feature templates include 12 kinds, according to the white cells lattice and black list of every kind of feature templates
The layout of first lattice obtains the algorithm of 12 kinds of image feature values, by by 12 kinds of feature templates respectively in the gray-scale map, gradient width
Moved in value figure, histogram of gradients by default step-length, extract substantial amounts of characteristics of image, and it is corresponding to calculate each characteristics of image
Characteristic value.
A kind of 4. object detecting method by feature templates construction feature pond according to claim 3, it is characterised in that:
Described image characteristic value subtracts the pixel value sum in the black unit lattice for the pixel value sum in the white cells lattice.
A kind of 5. object detecting method by feature templates construction feature pond according to claim 1, it is characterised in that:
In described step (30), refer to the feature of background where the characteristic value of each characteristics of image in the feature pool and object
Value is compared, and special as optimal object detection using maximum characteristics of image is distinguished with the characteristic value of background where object
Sign.
A kind of 6. object detecting method by feature templates construction feature pond according to claim 1, it is characterised in that:
In described step (40), refer to carry out structure object detection classifier using the optimal object detection feature, by this
Object detection classifier carries out object detection.
A kind of 7. object detecting system by feature templates construction feature pond, it is characterised in that including:
Image pre-processing module, for handling original image, obtain gray-scale map, gradient magnitude figure, histogram of gradients;
Feature pool builds module, its feature templates by setting more than one different grids, and the feature templates are distinguished
Moved in the gray-scale map, gradient magnitude figure, histogram of gradients, obtain the feature pool of characteristics of image;
Features training module, its use adaboost algorithms the characteristics of image in the feature pool is trained to obtain it is optimal
Object detection feature;
Object detection module, it carries out object detection according to the optimal object detection feature.
A kind of 8. object detecting system by feature templates construction feature pond according to claim 7, it is characterised in that:
In described feature pool structure module, the feature templates include 12 kinds, according to white cells lattice of every kind of feature templates and black
The layout of color element lattice obtains the algorithm of 12 kinds of image feature values, by by 12 kinds of feature templates respectively in the gray-scale map, ladder
Moved in degree amplitude figure, histogram of gradients by default step-length, extract substantial amounts of characteristics of image, and calculate each characteristics of image
Corresponding characteristic value.
A kind of 9. object detecting system by feature templates construction feature pond according to claim 7, it is characterised in that:
In described features training module, refer to background where the characteristic value of each characteristics of image in the feature pool and object
Characteristic value is compared, and distinguishes maximum characteristics of image as optimal object detection using with the characteristic value of background where object
Feature.
10. a kind of object detecting system by feature templates construction feature pond according to claim 7, its feature exists
In:In described object detection module, refer to carry out structure object detection classifier using the optimal object detection feature,
Object detection is carried out by the object detection classifier.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710426358.4A CN107358240A (en) | 2017-06-08 | 2017-06-08 | A kind of object detecting method and system by feature templates construction feature pond |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710426358.4A CN107358240A (en) | 2017-06-08 | 2017-06-08 | A kind of object detecting method and system by feature templates construction feature pond |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107358240A true CN107358240A (en) | 2017-11-17 |
Family
ID=60273670
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710426358.4A Pending CN107358240A (en) | 2017-06-08 | 2017-06-08 | A kind of object detecting method and system by feature templates construction feature pond |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107358240A (en) |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102663363A (en) * | 2012-04-09 | 2012-09-12 | 中国科学院光电技术研究所 | Human detection method in single frame image |
CN103605977A (en) * | 2013-11-05 | 2014-02-26 | 奇瑞汽车股份有限公司 | Extracting method of lane line and device thereof |
CN103778431A (en) * | 2013-12-30 | 2014-05-07 | 温州医科大学 | Medical image characteristic extracting and identifying system based on two-directional grid complexity measurement |
-
2017
- 2017-06-08 CN CN201710426358.4A patent/CN107358240A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102663363A (en) * | 2012-04-09 | 2012-09-12 | 中国科学院光电技术研究所 | Human detection method in single frame image |
CN103605977A (en) * | 2013-11-05 | 2014-02-26 | 奇瑞汽车股份有限公司 | Extracting method of lane line and device thereof |
CN103778431A (en) * | 2013-12-30 | 2014-05-07 | 温州医科大学 | Medical image characteristic extracting and identifying system based on two-directional grid complexity measurement |
Non-Patent Citations (1)
Title |
---|
查宇飞等: "视频目标跟踪方法", 《国防工业出版社》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106446930B (en) | Robot operative scenario recognition methods based on deep layer convolutional neural networks | |
CN103886308B (en) | A kind of pedestrian detection method of use converging channels feature and soft cascade grader | |
CN107945153A (en) | A kind of road surface crack detection method based on deep learning | |
Nie et al. | Pavement distress detection based on transfer learning | |
CN103218831B (en) | A kind of video frequency motion target classifying identification method based on profile constraint | |
CN106529448A (en) | Method for performing multi-visual-angle face detection by means of integral channel features | |
CN103049733B (en) | Method for detecting human face and human-face detection equipment | |
CN103605977A (en) | Extracting method of lane line and device thereof | |
CN107169953A (en) | Bridge concrete surface crack detection method based on HOG features | |
CN103632170A (en) | Pedestrian detection method and device based on characteristic combination | |
CN103310194A (en) | Method for detecting head and shoulders of pedestrian in video based on overhead pixel gradient direction | |
CN104156731A (en) | License plate recognition system based on artificial neural network and method | |
CN103093212A (en) | Method and device for clipping facial images based on face detection and face tracking | |
CN103048329A (en) | Pavement crack detecting method based on active contour model | |
CN110659550A (en) | Traffic sign recognition method, traffic sign recognition device, computer equipment and storage medium | |
CN102855758A (en) | Detection method for vehicle in breach of traffic rules | |
CN104978567A (en) | Vehicle detection method based on scenario classification | |
CN105868734A (en) | Power transmission line large-scale construction vehicle recognition method based on BOW image representation model | |
CN104915642B (en) | Front vehicles distance measuring method and device | |
CN103745197B (en) | A kind of detection method of license plate and device | |
CN106845458B (en) | Rapid traffic sign detection method based on nuclear overrun learning machine | |
CN106485273A (en) | A kind of method for detecting human face based on HOG feature and DNN grader | |
Masita et al. | Pedestrian detection using R-CNN object detector | |
CN105224911A (en) | A kind of various visual angles pedestrian detection method and system in real time | |
CN106940791A (en) | A kind of pedestrian detection method based on low-dimensional histograms of oriented gradients |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171117 |